Can Money Buy Success?: A Study of Team Payroll and Performance in the MLB

Shahriar Hasan, Thompson Rivers University, Canada

ABSTRACT

Using the last 16 years of winning percentages and pay-scales of the MLB teams, I investigated the relationship between team performance and payroll. I found strong evidence that during the regular season games, the payroll plays a significant positive role on the performance. Although the results of the individual team studies were not conclusive, the results were prominent when all teams are included in the data. Different variations of the pooled data were used to test the robustness of the outcome. I also found weak evidence that being a new team in the MLB may have negative effects on performance.

INTRODUCTION

In his 2002 eye-opening book Moneyball, Michael Lewis asked the question: “if the teams with the highest salaries consistently do well, why are the always one of the best teams with one of the lowest payrolls”? In 2001, when the Arizona Diamondbacks won the MLB crown being the 7th largest spender in the league and in only their 4th year of operations, it was understood that team payrolls could not be the only determinant of success. While a considerable amount of research has been done on the relationship between individual player salaries and respective performances, few studies have sought to examine the larger and more encompassing issue of team payrolls as the deciding factor of overall team performance. This paper, following the line of reasoning by Hall, Szymanski & Zimbalist (2002), attempts to establish a relationship between team payrolls and team performance in the Major League using 1992-2007 data. In 2005-06, the minimum salary for a player in the MLB was US$327,000 while the average was an astounding US$2,476,589. With that kind of money circulating, one would reasonably assume that the majority of the players are sufficiently motivated to give their best efforts. Furthermore, since the inception of the free agency system in 1976, players have had the freedom to move freely between clubs, thus giving one reason to believe that highest spenders would accumulate the most amount of talent. In a world of free movement and free information, teams with the highest payrolls should consistently win.

BACKGROUND

One of the earliest researchers in this field, Rottenberg (1956) focused on mostly why and how playing talent would be more or less equally distributed among teams given that teams would want to win closely contested games. Before the introduction of the free agency system, Scully (1974), in his seminal article, investigated the issue of individual players’ wages against their estimated marginal revenue product. He found the wages to be significantly lower than their respective MRPs. Zimbalist (1992) used a modified Scully method to dispense the worries about this inequality – he basically showed that the 1976 introduction of free agency system almost eliminated the monopsony powers of the team owners and brought the wages up to their MRP levels. Scully’s 1989 model have been scrutinized by others in different ways. Sommers & Quinton (1982) formalized a team revenue function and showed that each extra win contributes more marginal revenue in larger markets than smaller ones. This is a result that the Zimbalist (1992) study did not find. Before the publication of Zimbalist’s article, Scully came out with another path breaking work on pay and performance in baseball. In his 1989 book, he drew the conclusion that team revenues are directly related to the club’s

win percentage and that the size of the home market also positively affects team revenues. But Quirk & Fort (1999) argued that under free agency teams should get pretty much what they paid for. They looked at the correlation between the rank of regular season winning percentages and the rank of player payroll cost for the years of 1990 to 1997. Their findings were that this correlation was 0.509 in the and 0.135 in the . However, none of these were statistically significant and thus they concluded that payrolls could not explain the team performance in baseball. The 1992 Zimbalist study also lent support to this notion that team performance and payroll are not strongly correlated. Finally, Hall, Szymanski and Zimbalist (2002) used a 20 year data set to examine this link in both MLB and English Soccer. Additionally, they also tested for causality – whether the relationship runs from payroll to performance or vice versa. They were able to demonstrate that the correlation between performance and payroll increased significantly during the 1990s as compared to the 1980s, however, they failed to establish the direction of causality for the entire period. In a related study, Burger and Walters (2003) proved the existence of market size effect on expected team performance. They showed that ceteris paribus, baseball teams in the largest markets will value a given player six times more than those in the smallest. Also, within each market, achieving contending status can raise a player’s value significantly.

THE MODEL

For this paper, a definition of success has been adopted which does not extend in to the playoff season. It is assumed that the 162 games played during the regular season give us a more accurate picture of success of a team. The ranking of a team in regular season was not used as the indicator of success. Clearly illustrating the inherent problem associated with such an approach, the difference between a team ranked 5th and a team ranked 15th, is not necessarily by a factor of three. Thus, in terms of team spending, unless the 5th rank team were to spend three times the amount of the 15th rank team, analysis would always suggest a non-relation or not-so-strong relation. In this study, the percentage of wins during the regular season was chosen as the mark of success. On the determinant side, the team payroll is the lonely variable. A seasonally adjusted figure would not have done justice to the study since the increase in payment would not be accompanied by year to year increases in the winning percentage. Instead, the team payroll variable was standardized by dividing each team’s yearly payroll by the average of that year’s league payroll. Not only did this bring harmony in the analysis, but this aids the reader to see which teams were spending more/less than the average on a yearly basis.

Using the equation: Win Pctt = α + β Pay-scalet + εt, a relationship was examined between pay and performance in both the team and the overall (pooled) level. The pay-scale in the equation is the percentage of the actual team payroll compared to the average payroll for the overall league for the year. For both levels of study, a simple OLS was run to see whether β is significantly positive or not. A decision was made to add a dummy variable that covered the first three years of operations for the four newcomer teams – Colorado, Florida, Arizona and Tampa Bay. (The first two started operation in 1993 while the latter two in 1998). The reason for introducing the dummy variable was to allow for the effects of initial internal adjustments on the teams’ performances. Besides the addition of the four teams as mentioned above, the MLB also allowed two teams to relocate – the California Angels became Anaheim Angels in 1997 and the Montreal Expos relocated to Washington DC as the Nationals in 2003. It was assumed that these changes did not affect the team performance drastically, as both teams kept their management and players pretty much unchanged during the transition. While running the pooled level tests, the effect of the new club status was examined to see if it has any negative impact on the winning percentage or not. Using a proxy variable for the new teams for their first three years of operation, the following equation

Win Pctt = α + β Pay-scalet + λ Dummyt + εt, was used to test whether β is significantly positive and if λ, significantly negative.

DATA COVERAGE

The relationship between payroll and performance was investigated, using both pooled and individual team data for during the period 1992 to 2007. This particular period was selected so that the market uncertainties before the 1994-95 MLBPA strike could be minimized over the entire period. The data is easily available through a multitude of electronic and published sources. For team level studies, pay-scale averages for each team were computed to give a historical snapshot of who had been the big spenders during the last 16 years. On top of this, an assumption was made that spending more than 100% of the average payroll should bring 50% or more wins for the teams. Thus, if a team spent more than 100% but won less than 50%, it was termed as an underachievement. Similarly if a team spending less than 100% won more than 50%, it was considered to an overachievement. When using the pooled data to run the regressions, the entire league was divided into two groups - the first with payroll figures being greater than the 100% of the average payroll; and the second with less than 100%. A similar division was also performed from the winning perspective – the first group had a 50% or greater winning percentage, and the second group of less than 50%. This division allowed an objective assessment of whether the pay-scale effect is uniformly applicable at all levels. To be a little more precise, the above mentioned divisions were further analyzed and sub-divided into three groups: on the pay-scale basis, the first group had 120% or more, the second one had 80 to 120% and the third group had less than 80% of the pay-scale. In terms of wins, the first group had a 55% or more winning record; the second group from 45 to 55%; and the third, a less than 45% winning record. For all these divisions, regressions were run both with and without inclusion of the dummy variable as mentioned above. Following Zimbalist (2002), the above regressions were also run for every year for the pooled data. Lastly, consistent with the other studies, all the above mentioned tests were run separately on the two leagues – American and National. Outcomes indicated minimal differences between the two, showing instead that they seem to operate pretty much in unison. Figures and results have not been reported in this paper.

THE RESULTS

Results of the study are presented in two sections: first, the individual team results and secondly, the results for the pooled data. Tables 1 to 4 refer to the individual team data while the remaining tables relate to the pooled data. Table 1 details the averages of the winning percentages of each team over the sample period along with their average pay-scales expressed in percentages of the yearly average of the league. Additionally, it shows the growth rate of the team payroll during this time. This is the average of the year to year growth in their payrolls. The last two columns show how many years did the team won more than 50% while they spent less than 100% pay-scale (overachievement) and the number of years they won less than 50% while they spent more than 100% pay-scale (underachievement). Oakland Athletics has the highest over achievement (8 times) whereas Baltimore Orioles has the highest number of under achievement (9 times). Table 2 shows the average winning percentage and average payrolls of the league, divided into the two groups. We see that the group spending more than 100% on average has a 52.01% winning percentage, as opposed to the under- spenders’ 47.76%. When divided according to winning percentage, we see that teams with 50% or more winning percentage spent an average of 115.78% of the league average. This is in sharp contrast to the other group (less than 50% winners) that spent 85.7% of the league average. Both these observations support the hypothesis that team spending improves winning percentage. Table 3 describes the results of the individual team regressions. Teams shown in bold letters produced statistically significant coefficients (7 of them at 5% level and 2 of them at 10% level). At best, the results could be termed as inconclusive with 70% of the teams showing no relationship between winning percentage and expenditures. This is precisely the reason why panel data was used to test the hypothesis. The results of the entire panel data regression are shown in Table 4. The first row shows the results when only the pay-scale is used as the determinant. The coefficient is highly significant and the adjusted R square is respectable.

These results perfectly complement the hypothesis that payroll affects the winning percentage in a positive manner. When the same regression was run with a proxy variable added to capture the effect of being a new team, the coefficient turns out to be negative and significant at the 10% level. This was an unusually strong outcome that was not expected. The inclusion of the dummy also raised the adjusted R square indicating that being a new team may have some explanatory power on the winning performance of a team. In order to establish the robustness of the strong results from Table 4, the same regressions were tried on four different grouping schemes. First, the league was divided according to the pay-scale. Within this division, two variations were tested – divisions into two and three groups. Table 5 shows the results of the two group division and Table 6 the three group divisions. The league was then divided in terms of winning percentage – again, both in two and three groups. Table 7 shows the two group results, and table 8 the three group results. For all four divisions, regressions including and excluding the newcomer proxies for each variation were run. For the teams representing the lowest third of the pay-scale, it appeared that there is a weak relationship between pay and performance. With the exception of this, every other group demonstrated strong positive relationships between pay and performance. Finally, following Zimbalist et. al (2002), regressions were run for each of the sample years and are shown in Table 9. Because the pay-scale was used (as opposed to the payroll), the results were different than that of Zimbalist study. However, in keeping with their findings, strong evidence was found in support of performance being significantly affected by team payrolls, even on a cross sectional basis. During 1998 and 1999, this relationship was quite strong as evidenced by the high adjusted R squares. During the last four years, this relationship seems to have steadied as evidence by high t-statistics and moderate adjusted R squares.

CONCLUSIONS

This study used only the regular season game results as outcomes of the team payroll. Also, unlike other studies, the pay-scale was used instead of the seasonally adjusted monetary units to depict the predictor. Despite recent evidence that payroll may not be an important factor in team success in the Major League Baseball, sufficient evidence was found that regular season outcomes are highly influenced by how much the teams spend on their players. No matter from which angle it was examined, there exists ample evidence of this. An additional outcome of the study was the mixed evidence found indicating that newcomers in the MLB may have to wait a few years to fully achieve their respective levels of success. The results are in sharp contrast to the Zimbalist (1992) and the Quirk & Fort (1999) outcomes. Although the success stories of the Oakland Athletics and the failures of the may induce the casual observer to conclude that baseball game outcomes are not money-driven, sufficient proof has been compiled to consider these examples as outliers, at best. Compared to other professional sports, baseball has a much higher degree of uncertainty, as evidenced in the highest winning percentage of around 60% only (). Yet the fact remains that teams who spend the most will be rewarded positively at least during the regular season.

Table 1: Individual Team Averages 1992-2007 Team Average Average Payroll Over Under Winning % Payroll Growth Achieve Achieve Atlanta Braves 59.96% 135.95% 4.53% 0 1 Baltimore Orioles 48.39% 116.95% 9.71% 2 9 Boston Red Sox 54.22% 140.67% 7.31% 1 4 Los Angeles Angels 50.49% 102.18% 7.27% 5 3 Chicago Cubs 47.91% 113.08% 6.56% 1 7 Chicago White Sox 52.57% 103.12% 10.81% 5 3 49.41% 94.79% 6.10% 2 2 54.06% 98.997% 16.10% 3 1 43.90% 88.86% 9.87% 0 3

Houston Astros 53.30% 96.89% 13.37% 6 1 Kansas City Royals 43.85% 75.27% 7.77% 2 1 Los Angeles Dodgers 51.51% 131.38% 5.13% 1 3 45.80% 69.49% 6.99% 2 0 Minnesota Twins 48.93% 69.38% 6.80% 7 0 Washington Nationals 48.32% 53.15% 14.23% 7 0 49.68% 127.19% 6.29% 0 5 New York Yankees 58.97% 186.29% 9.44% 0 1 Oakland Athletics 52.33% 79.29% 4.25% 8 3 49.12% 94.79% 8.13% 3 0 Pittsburg Pirates 44.91% 60.18% 7.27% 0 0 San Diego Padres 48.47% 80.91% 6.43% 6 0 San Francisco Giants 52.35% 109.39% 4.76% 5 6 51.20% 110.60% 9.33% 0 4 St. Louis Cardinals 52.91% 109.49% 6.70% 2 2 Texas Rangers 49.39% 116.59% 5.73% 1 6 Toronto Blue Jays 50.07% 108.45% 5.86% 4 5 Colorado Rockies 47.10% 88.80% 16.47% 2 3 Florida Marlins 47.01% 63.75% 13.55% 3 0 Arizona Diamondbacks 50.49% 108.24% 10.74% 1 1 Tampa Bay Devil Rays 39.89% 58.06% 4.40% 0 1

Table 2: Group Statistics 2 Groups Group Average Average Over Under Winning % Payroll Achieve Achieve Over 100% 52.01% 121.30% 1.87 4 Under 100% 47.76% 78.71% 3.4 1 Over 50% 53.17% 115.78% 2.93 2.71 Under 50% 47.01% 85.70% 2.38 2.3

Table 3: Team-wise Regressions Team Payroll Adjusted Coefficient R-square Atlanta Braves 0.178** 0.295 Baltimore Orioles 0.3704** 0.5154 Boston Red Sox 0.0722 0.0696 Los Angeles Angels 0.149** 0.2599 Chicago Cubs 0.103 -0.0285 Chicago White Sox -0.0063 -0.07 Cincinnati Reds 0.0704 0.0436 Cleveland Indians 0.0806** 0.1684 Detroit Tigers 0.1604** 0.1968 Houston Astros 0.0776 0.0046 Kansas City Royals 0.1164** 0.2689 Los Angeles Dodgers -0.025 -0.0601 Milwaukee Brewers 0.1132 0.0649 Minnesota Twins 0.0959 0.0152 Washington Nationals 0.1103 -0.0155 New York Mets 0.0146 -0.0679 New York Yankees 0.0329 0.0324 Oakland Athletics -0.0633 -0.0147 Philadelphia Phillies 0.1087 0.087 Pittsburg Pirates 0.0871 0.0815 San Diego Padres .2032** 0.2395

San Francisco Giants -0.1008 -0.043 Seattle Mariners 0.1757 0.0401 St. Louis Cardinals 0.1769* 0.1174 Texas Rangers 0.002 -0.0712 Toronto Blue Jays 0.0247 -0.0517 Colorado Rockies 0.0544 0.0366 Florida Marlins 0.0737 0.033 Arizona Diamondbacks 0.1551* 0.2173 Tampa Bay Devil Rays 0.0237 -0.0648

Table 4: Results of Pooled Regression Overall League Regression Data Specification Size Variable Coefficient t-statistics Adj. R Square

No Dummy 466 Pay-scale 0.0864 10.8813 0.2016

Start-up Dummy 466 Pay-scale 0.085 10.6835 0.2056 Dummy -0.0296 -1.8264

Table 5: Results of Pooled Regression League divided in to two groups according to payroll. Group & Data Specification Size Variable Coefficient t-statistics Adj.R square Above Avg. Payroll No Dummy 227 Pay-scale 0.0762 4.5624 0.0806 Below Avg. Payroll No Dummy 239 Pay-scale 0.0785 3.7539 0.0521 Above Avg. Payroll With Dummy 227 Pay-scale 0.0762 4.5553 0.0766 Dummy -0.0073 -0.1913 Below Avg. Payroll 239 Pay-scale 0.0773 3.7129 0.0629 With Dummy Dummy -0.034 -1.9357

Table 6: Results of Pooled Regressions League divided in to three groups according to payroll. Group & Data Specification Size Variable Coefficient t-statistics Adj.R square Highest 3rd payroll No Dummy 131 Pay-scale 0.0676 3.2467 0.0684 Medium 3rd payroll No Dummy 192 Pay-scale 0.1126 2.9142 0.0377 Lowest 3rd payroll No Dummy 143 Pay-scale 0.0633 1.7037 0.0132 Highest 3rd payroll Pay-scale 0.0677 3.2424 With Dummy 131 Dummy 0.0315 0.7084 0.0648 Medium 3rd payroll Pay-scale 0.1119 2.8931 With Dummy 192 Dummy -0.0302 -0.8125 0.036 Lowest 3rd payroll Pay-scale 0.0719 1.9435 No Dummy 143 Dummy -0.0403 -1.9805 0.0332

Table 7: Results of Pooled Regressions League divided in to two groups according to winning percentage Group & Data Specification Size Variable Coefficient t-statistics Adj.R square Above 50% No Dummy 230 Pay-scale 0.0329 4.6764 0.0835 Below 50% No Dummy 236 Pay-scale 0.0221 2.7921 0.0281

Above 50% With Dummy 230 Pay-scale 0.0328 4.6671 0.0819 Dummy -0.0152 -0.77 Below 50% With Dummy 236 Pay-scale 0.0211 2.6451 0.0298 Dummy -0.0135 -1.1889

Table 8: Results of Pooled Regression League divided in to three groups according to winning percentage. Group & Data Specification Size Variable Coefficient t-statistics Adj.R square Above 55% win No Dummy 117 Pay-scale 0.0135 1.9726 0.0243 45 to 55% win No Dummy 229 Pay-scale 0.024 3.7947 0.0555 Below 45% win No Dummy 120 Pay-scale 0.0183 2.0642 0.0267 Above 55% win Pay-scale 0.0133 1.9294 With Dummy 117 Dummy 0.0225 0.7308 0.0264 45 to 55% win Pay-scale 0.0239 3.7806 With Dummy 229 Dummy -0.0013 -0.0989 0.0514 Below 45% win Pay-scale 0.0192 2.1569 No Dummy 120 Dummy 0.0099 0.9819 0.0264

Table 9: Results of Pooled Regression Year-wise results Year Beta t-stats Adj. R square n 1993 0.076 1.65 0.0599 28 1994 0.1053 2.2607 0.1321 28 1995 0.0732 1.5253 0.0468 28 1996 0.1142 3.6396 0.312 28 1997 0.082 2.7164 0.1911 28 1998 0.1521 4.9685 0.4496 30 1999 0.1208 5.2569 0.4787 30 2000 0.0505 1.7941 0.0711 30 2001 0.0677 1.7811 0.0697 30 2002 0.1109 2.6228 0.1685 30 2003 0.0875 2.4416 0.1461 30 2004 0.0943 3.3798 0.2644 30 2005 0.0704 3.0027 0.2166 30 2006 0.0803 3.3704 0.2632 30 2007 0.0679 3.0307 0.2201 30

REFERENCES

Brugging, T. H., & Rose, D. R. (1990). Financial Restraint in the Free Agent Labor Market for Major League Baseball: Players Look at Strike Three. Southern Economic Journal, 56, 1029-1043. Burger, J. D., & Walters, J. K. (2003). Market Size, Pay and Performance: A General Model and Application to Major League Baseball. Journal of Sports Economics, 4(2), 108-125. Daly, G. (1992). The Baseball Players’ Market Revisited. In P. Sommers (Ed.), Diamonds are Forever: The Business of Baseball. Washington, DC: Brookings Institution. Eckard, E. (2001). Free Agency, Competitive Balance and Diminishing Returns to Pennant Contention. Economic Inquiry. 39(3), 430-443. Fort, R., & Quirk, J. (1995). Cross Subsidization, Incentives and Outcomes in Professional Team Sports Leagues. Journal of Economic Literature, 33(3), 1265-1299. Hall, S., Szymanski, S., & Zimbalist, A. S. (2002). Testing Causality Between Team Performance and Payroll: The Cases of Major Leagues Baseball and English Soccer. Journal of Sports Economics. 3(2), 149-168.

Hylan, T., Lage, M., & Treglia, M. (1996). The Coase Theorem, Free Agency, and Major League Baseball: A Panel Study of Mobility From 1961 to 1992. Southern Economic Journal. 62, 1029-1042. Kahn, L. M. (2000). The Sports Business as a Labour Market Laboratory. Journal of Economic Perspectives. 14(3), 75-94. Levin, R. C., Mitchell, G. J., Volcker, P. A., & Will, G. F. (2000). The Report of the Independent Members of the Commissioner’s Blue Ribbon Panel on Baseball Economics. New York: Major League Baseball. Retrieved from http://www.mlb.com/mlb/downloads/blue_ribbon.pdf. Lewis, M. (2003). Moneyball: The Art of Winning an Unfair Game. New York: Norton. Quirk, J., & Fort, R. (1999). Hard Ball: The Abuse of Power in Pro Team Sports. Princeton, NJ: Princeton University Press. Raimondo, H. (1983). Free Agents’ Impact on the Labour Market for Baseball Players. Journal of Labor Research, 4, 183-193. Rottenberg, S. (1956). The Baseball Players’ Labour Market. Journal of Political Economy, 64, 243-256. Scully, G. W. (1974). Pay and Performance in Major League Baseball. American Economic Review, 64, 915-930. Scully, G. W. (1989). The Business of Major League Baseball. Chicago: University of Chicago Press. Sommers, P. M., & Quinton, N. (1982). Pay and Performance in Baseball: The Case of the First Family of Free Agents. Journal of Human Resources, 17, 426-436. Vrooman, J. (1997). A Unified Theory of Capital and Labour Markets in Major League Baseball. Southern Economic Journal, 63, 594-619. Zimbalist, A. S. (1992). Salaries and Performance: Beyond the Scully Model. In P. Sommers (Ed.), Diamonds are Forever: The Business of Baseball (pp. 109-133). Washington DC: Brookings Institution. Zimbalist, A. S. (2001). Competitive Balance in Major League Baseball. Milken Institute Review, 3(1), 54-64. Zimbalist, A. S. (2002). Competitive Balance in Sports Leagues: An Introduction. Journal of Sports Economics, 3(2), 111-121.